잠재적 의미 분석을 이용한 정보 검색 = Information Retrieval using Latent Semantic Analysis

다국어 초록 (Multilingual Abstract)

Most systems that manage distributed information have difficulties in retrieving relevant information for they are not able to include the exact semantics of retrieval queries that users request. In this paper, we proposed automatic query expansion based on term distribution which reflect semantic of retrival term to enhance the performance of information retrival and we evaluated the performance.
We use LSI(Latent Semantic Indexing)'s SVD(Singular Value Dacomposition) technique to measure the term distribution which appear similar to query. And we measure the similarity to compared numerical value with query terms. Also we researched the method for setting the threshold according to area distribution in other to choose additional terms automatically. Finally, we applied the proposed method in this paper to two experimental DB's to evaluate the performance.

목차 (Table of Contents)

• 1. 서 론
• 2. 관련연구
• 2.1 용어 분류 기법
• 2.2 관련성 피드백 기법
• 2.3 용어 발생 확률 기법
• 3. LSI(Latent Semantic Indexing)을 이용한 정보 검색
• 3.1 LSI의 기본개념
• 3.2 SVD를 이용한 용어의 구조 분석
• 4. 정보 검색 시스템 구성
• 4.1 용어 추출과정
• 4.2 특이치 분해(SVD)과정
• 4.3 질의어 확장 과정
• 5. 실험 및 평가
• 5.1 실험 모델 구성
• 5.2 측정 요소
• 5.3 정확도(precision)변동률
• 5.4 재현도(recall) 변동률
• 5.5 유효성 척도(E,effectiveness measure)변동률과 개선률
• 5.6 평균 정확도(precison)를 이용한 종합 평가
• 6. 결 론